Contents lists available at ScienceDirect

Marine Pollution Bulletin

journal homepage: www.elsevier.com/locate/marpolbul

Baseline

Nearshore sea surface macro marine debris in Maui County, Hawaii:Distribution, drivers, and polymer compositionJens J. Curriea,⁎, Stephanie H. Stacka, Kayla C. Brignacb, Jennifer M. Lynchc

a Pacific Whale Foundation, Maui, HI, USAb School of Ocean, Earth Science, and Technology, University of Hawaii at Manoa, Honolulu, HI, USAc Chemical Sciences Division, National Institute of Standards and Technology, Pacific University, Kaneohe, HI, USA

A R T I C L E I N F O

Keywords:Plastic pollutionMonitoringAccumulationPolyethyleneOcean surface currents

A B S T R A C T

Located within the subtropical convergence zone, the Hawaiian archipelago is subject to high debris loads. Thispaper represents the first study to determine the spatial and temporal trends of floating macro debris quantitiesand polymer composition within Maui County waters. Ocean surveys were conducted from 2013 to 2017 andcollected 2095 debris items of which 90% were plastic. Attempts to categorize items by source resulted in only6% likely from land, 12% from ocean-based sources, 50% from either land or ocean, and 32% from unknownsources. Results found a multi-step process for debris accumulation, with temporal trends linked to survey dayand year and spatial trends linked to ocean processes. High- and low-density polyethylene and polypropyleneaccounted for the majority of polymer types. The results of this study demonstrate minimal debris in Mauioriginates from land/local sources, and the importance of baseline data to guide further research and mitigationmeasures.

Marine debris poses a considerable threat to marine life, biodi-versity, and ecosystems (Sheavly and Register, 2007; Galloway et al.,2017) and has been identified as a stressor for a variety of marine life(Moore, 2008; Currie et al., 2017). Marine debris can be classified intothree categories describing its likely source: land, ocean, and “general”,which encompasses both or either land and ocean, as described by Ribicet al. (2012). Previous research has identified ocean-based debris as theprimary source of Hawaiian marine debris (Donohue et al., 2001), withproportionately higher ocean-based debris when compared to otherregions in the Pacific (Ribic et al., 2012). Once reliable data on wheremarine debris originates and how it is introduced into the marine en-vironment is available, targeted efforts to stop this problem at thesource can be implemented.

Current knowledge of ocean currents in the North Pacific suggeststhree high-density areas of debris accumulation based on convergencezones (Wakata and Sugimori, 1990; Kubota, 1994; Van Sebille, 2015).One such zone is located just north of the Hawaiian Islands, and hasbeen found to accumulate debris (Donohue et al., 2001; Pichel et al.,2007; Goldstein et al., 2013) (Appendix Fig. 1). The origins of debrisnorth of Hawaii varies greatly, and the resulting accumulation is theresult of multi-step processes starting with the Ekman convergencezone, transport via the geostrophic currents, and finally Ekman drift(Kubota, 1994). Marine debris accumulating north of the Hawaiian

archipelago can travel through various marine ecosystems includingcoastlines, remote islands, the open ocean, and subtropical gyres(Derraik, 2002; Barnes et al., 2009). Some work has been conducted todocument the rates and process of marine debris accumulation in theNorthwestern Hawaiian Islands (Kubota, 1994; Donohue et al., 2001;Dameron et al., 2007; Pichel et al., 2007), but these efforts have beenminimal and rates are likely out of date. Further, this work does notinclude the coastal waters of the Main Hawaiian Islands.

Currie et al. (2017) presented the first study quantifying the amountand type of marine debris in the nearshore waters of Maui county andrelated this to cetacean distribution to identify areas where marinedebris ingestion or entanglement may present a high risk to marinemammals. Ingestion and entanglement of marine debris by biota hasbeen well documented (Kühn et al., 2015), and effects of plastics onmarine life are polymer dependent (Rochman et al., 2013). The currentstudy expands the previous study (Currie et al., 2017) by performingpolymer identification and statistical models to find local and ocean-wide variables that explain the accumulation of plastic marine debrisfloating in Maui County's nearshore waters.

Plastic marine debris is comprised of many different polymers thathave specific chemical compositions defining their physical and che-mical properties, leading to different environmental fate and effects.Polymer composition of marine debris items will affect vertical

https://doi.org/10.1016/j.marpolbul.2018.11.026Received 15 August 2018; Received in revised form 7 November 2018; Accepted 13 November 2018

⁎ Corresponding author.E-mail address: research@pacificwhale.org (J.J. Currie).

Marine Pollution Bulletin 138 (2019) 70–83

0025-326X/ © 2018 Published by Elsevier Ltd.

T

stratification in the water column and influence interactions withmarine organisms (Jung et al., 2018). Debris collected from the surfaceis expected to consist of floating low-density polymers such as poly-ethylene (PE) and polypropylene (PP) as opposed to sinking high-den-sity polymers such as polyethylene terephthalate (PET) (Jung et al.,2018). As plastic marine debris is a growing problem in Maui County, aneed for identifying polymer type is crucial for understanding the be-havior of debris in the ocean environment and to know which polymerswhales, dolphins, and other marine life will be exposed to in order toreduce their impact.

Blickley et al. (2016) monitored shoreline debris in Maui and founddebris loads were linked to ocean based processes such as winds andcurrents as well as beach exposure and location. At a single siteshoreline debris accumulation rates were as high as 1460 items per daywithin a 100 meter section, which was largely attributed to ocean basedsources (Blickley et al., 2016). A clearer understanding of nearshoredebris loads within the waters surrounding the Main Hawaiian Islandsshould be determine to supplement Blickley et al. (2016) and allow fora better understanding of which mitigation measures would be mosteffective for Hawaii. This study quantified macro (> 1 cm diameter)marine debris floating in the nearshore waters of Maui County as acomplement to existing shoreline surveys. The objectives of this studywere to: (1) identify factors influencing marine debris accumulation;and (2) characterize debris type including polymer composition, source(ocean, land, or general), and amount of marine debris within the studyarea.

The nearshore waters of Maui county that made up the study area(center: 20.73623°N, 156.69085°W) are semi-enclosed by the islands ofMaui, Molokai, Lanai, and Kahoolawe and located within the HawaiianIslands Humpback Whale National Marine Sanctuary. The channelsbetween the islands are akin to drowned land bridges that once con-nected the surrounding islands. The study area consists predominantlyof nearshore habitats with gently sloping shoreline gradients that ex-tend to more complex bathymetry of seamounts and ridgelines (Grigget al., 2002). The majority of the study area consists of drowned reeffeatures and sandy basins with a depth of < 200 m; however, someareas south of Lana'i reach depths up to ≈600 m.

The detailed methods of data collection for this study are previouslydescribed in Currie et al. (2017), with a brief synopsis provided here.Line transects, separated by one nautical mile (1.85 km), were con-ducted within the leeward waters, up to 18 km from the coasts, of thefour islands (Maui, Molokai, Lanai, and Kahoolawe) that comprise MauiCounty, Hawaii and covered 1004 km2. Line transects were followedusing the research vessel's onboard GPS. The starting point of eachsurvey was chosen randomly at the beginning of each survey day. Thetransect lines, as followed using the onboard GPS, are presented inFig. 1, with survey effort shown in Fig. 2. If debris was sighted during atransect, the transect was paused while the vessel changed course topick up the debris. Once the debris was removed and documented, thevessel navigated back to the transect line and effort was resumed fromthe pause position. To ensure there were no missed occurrences ofdebris, all sightings of marine debris, regardless if on transect or tra-velling between transects, were recorded and used in subsequent ana-lysis.

A minimum of one survey day/week was attempted and planned forthe day with the best weather forecast, to allow observers the bestconditions for visibility. If no suitable weather days (Beaufort andDouglas Sea States were > 3) were available, multiple surveys wereconducted during the next suitable weather window. From April 6,2013 to October 12, 2017, 767 line transect surveys for macro (> 1 cmdiameter) floating debris were completed over 260 days from an 8 mengine-powered research vessel. The collection of debris for this projectwas done in conjunction with a systematic line transect study forodonotocetes, and it is important to note that despite conducting linetransect surveys, distance sampling procedures were not followed forthe debris items collected. Therefore, no effective transect width could

be calculated and the survey width was limited to the sighting distanceof the observer. As such, the results presented here represent presence-only sightings, which have not been corrected for detectability. Thetransect surveys ensured sufficient coverage of the survey area, but notadhering to systematic survey methods allowed for the highest numberof debris items to be collected and was deemed most appropriate forthis study.

All marine debris location data were imported into ArcGIS 10.6(Environmental Systems Research Institute, 2012) and mapped with theWorld Mercator projection, using the WGS 1984 datum. To determinespatial trends in debris quantities over the entire duration of the studyperiod, the study area was divided into 1004 grid cells each with anarea of 1 km2 (1 km × 1 km). Each grid cell was classified by the countof debris items occurring in that cell and the total survey distance (km)travelled within the boundaries of that cell from April 6, 2013 to Oc-tober 12, 2017. Quantities of marine debris were summarized per gridcell by dividing the sum of debris counts by the sum of survey effort(km) within each grid cell, resulting in final units of number of debrisitems/km/grid cell, which henceforth will be referred to as spatialquantities. Grid cells with no survey effort were dropped and not dis-played in the final spatial trend map.

All floating macro debris (> 1 cm diameter) items within sightingdistance were recorded during the survey period. As such, results pre-sented here likely represent an underestimation of all debris items be-cause smaller items, particularly micro and nano debris items, were notsampled. Observations were undertaken by two experienced observersstationed on the port and starboard sides of the vessel, as well as theboat operator who was stationed at the helm using a continuous scan-ning methodology (Mann, 1999) by naked-eye or reticle binoculars(Bushnell 7 × 50), while a fourth person acted as a data recorder. Noelevated observation platform was used and, as such, each observer'sfeet were standing ≈28 cm above the waterline. To ensure minimalinfluence of weather on detectability of debris, surveys were onlyconducted in the absence of rain and when Beaufort and Douglas SeaStates were ≤3. However, there is likely some un-corrected influence ofweather on the detectability of debris that should be acknowledged,and the debris counts recorded for this study may represent an under-estimate of the true count. When a piece of debris was sighted, the itemwas collected (if size and conditions allowed) and latitude, longitude,type of material, and percent of organism coverage (biofouling) on thedebris item were recorded. Percent organism coverage was determinedby visual inspection of the debris item and estimating the proportion ofbiofouling with respect to total surface area (See Appendix Fig. 2). If theitem could not be collected given the feasibility of removing it from thewater, it was photographed and recorded but left in the ocean.

To be consistent with the national debris monitoring program, alldebris classification was based on standardized source categories es-tablished by the United States Environmental Protection Agency(Escardó-Boomsma et al., 1995) and detailed in Ribic et al. (2012).Debris was divided based on the type of debris: plastic, metal, glass,rubber, clothing/fabric, processed lumber; and probable source cate-gory: ocean, land, general, or unknown. Ocean-based debris related toocean recreation and commercial fishing; land-based debris related toland-based recreation and activities; general-sourced debris related toitems that could originate from either ocean- or land-based sources(Ribic et al., 2012); and unknown-sourced debris consisted of debrisfragments that could not be identified and therefore could not be re-liably placed in a source category. The probable source categories(ocean, land, and general) were adapted from Ribic et al. (2012) andthe debris item division used in this paper is presented in AppendixTable 1. It should be noted that debris was classified as best as possiblein the most likely source category, but the potential for overlap betweencategories may exist and should be considered when interpreting re-sults.

To gain a better understanding of the type of plastic debris that wascollected throughout the survey period, all plastic debris items were

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

71

subclassified into nine main categories: foam fragments, food packagingfragment, net/rope fragments, plastic fragments, plastic bottles, plasticbags, buoys, jugs and other. Proportions of debris within these ninecategories were summarized and presented in the results.

To help identify what items were commonly found throughout thesurvey period, all intact items that were recorded a minimum of 15times were further subclassified into 14 categories regardless of debris

type: aluminum cans, balloons, beach toys, bottle caps, buckets, buoys,cups, fishing gear, food containers, food wrappers, jugs, nets/ropes,plastic bags, and plastic bottles. Proportions of debris within these 14categories were summarized and presented in the results.

Photos of debris items were visually inspected for writing or char-acters that may indicate country of origin (See examples in AppendixFig. 3). If non-English writing was present, country of origin was

Fig. 1. Map depicting study area and line transects surveyed within Maui County, Hawaii from April 6, 2013 to October 12, 2017 and overlaid with the HawaiiHumpback Whale National Marine Sanctuary (HIHWNMS).Note: The study area covers approximately ~20% of the HIHWNMS.

Fig. 2. Map showing (A) survey effort, and (B) marine debris spatial quantities (items/km of survey effort/grid cell) between April 6, 2013 and October 12, 2017within Maui County, Hawaii.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

72

assumed based upon the language displayed on the debris item.To assess sources and composition, results were summarized for

each survey year by dividing yearly sum of debris counts within each ofthe four debris source categories (ocean, land, general, unknown) bythe sum of the yearly survey effort (km). This resulted in final units ofdebris count/km/year. A two way ANOVA was used to test for differ-ences in yearly quantities across the source categories and year. Toaccount for unequal sample sizes within the independent variables, aType III sums of squares was employed in our ANOVA using the Anova() and aov() functions in R (R Core Team, 2017; Fox and Weisberg,2011). Before conducting the statistical analyses, the lack of normalitywas addressed by log transforming the data prior to analyses (Zar,1984).

Debris located within 50 m of each other were classified as a clusterof debris and considered an indicator of localized high accumulation.The mean number of debris items per cluster was calculated for eachyear to determine if cluster concentration varied with time.

All hard plastics collected from April 19, 2017 to July 11, 2017 wereanalyzed with a PerkinElmer attenuated total reflectance Fouriertransform infrared spectrometer Spectrum Two (Waltham, MA) (ATRFT-IR) for polymer identification using the method described in Junget al. (2018). Air-dried pieces were weighed to ± 1 g or for smallerpieces to ± 0.00001 g. Pieces were not cleaned prior to analysis, butwere cut with a razor blade when needed to expose a clean, smooth,and uncontaminated inner surface. Items that contained more than onepart (e.g., a bottle and a cap) were separated into multiple pieces foranalysis. All samples were assigned a color, opacity, and weatheringcode. Weathering codes were assigned visually as 1 = mild, 2 = mod-erate, and 3 = severe based on the intensity of square fracturing on thesurface of the sample with 1's having the least and 3's having the most(Appendix Fig. 4). Polymers were identified from spectra using ab-sorption bands, criteria, and the decision tree described in Jung et al.(2018). A float test in ethanol and deionized water solutions withdensities of 0.931 and 0.941 g/mL was performed as outlined in Junget al. (2018) on 21 unknown PE samples to differentiate between low-and high-density polyethylene (LDPE, HDPE).

Daily debris counts (items/day) were analyzed separately for land-based, ocean-based, general-source, and unknown-source as processesleading to changes in accumulation are likely to be different for eachcategory. However, there is value in understanding if all debris col-lected, regardless of source, is influenced more by land or ocean drivers.To determine this, two models of daily debris counts (items/day) weretested, one model using only land-based variables/processes and asecond model using only ocean based variables/processes, as describedbelow. The set of drivers (land or ocean) resulting in the lowest Akaike'sInformation Criterion (AIC) model was then considered to be the mostinfluential set of variables for describing overall debris trends withinthe study and presented in the results.

To account for potential nonlinear relationships between debriscounts and explanatory variables (Ribic et al., 2012), Generalized Ad-ditive Models (GAM) were constructed using the ‘mgcv’ package in R(Wood, 2017), using a gamma of 1.4 to avoid overfitting (Ribic et al.,2012; Wood, 2006). Daily debris counts (items/day) were modelled foreach source category and log-transformed for normality as a function ofsurvey variables, environmental variables, and process-based variables(partially adapted from Ribic et al., 2012 and explained below), with anoffset term for daily survey effort (km/day). Explanatory variables weretested for pairwise correlations using the stats package in R (R CoreTeam, 2017). To account for non-normality in variables, the Spearmancorrelation coefficient (rs) was used to assess correlations. If variableswere highly correlated (rs≥ 0.7) (Gonzalez-Suarez et al., 2013) onlythe variable that provided the lowest AIC value was retained.

To model temporal trends, a coded survey day (Ribic et al., 2012)was used, where 1 represented the first survey day (April 6, 2013) and1634 represented the last survey day (October 12, 2017). Variations inwithin-year deposition may be the result of human activity and/or

extreme weather events, which as described in Ribic et al. (2012), canconsistently be captured with month. Therefore, within-year trendswere modelled using month and between year variations using year asan explanatory variable.

The following variables were associated with each survey date usinghistorical National Data Buoy Center (NDBC) data (www.ndbc.noaa.gov): wind speed (km/h) and direction (degrees), peak gusts (m/s),wave height (m), dominant and average wave period (s), dominantwave direction (direction), sea level pressure (hPa), air temperature(Celsius), and sea surface temperature (Celsius). A total of 24 activebuoys are deployed within 500 km of the Hawaii islands chain, with themajority concentrated around the island of Oahu (National ClimaticData Center's (https://www.ncdc.noaa.gov), 2018). Two of these activebuoys were selected based on their proximity and location relative tothe study area and the metrics they recorded. The two data sourceswere evaluated independently for analysis of ocean-based debris, asvariables differed between the two sources. Only the source data setproviding the lower AIC model was presented in the final results. Datawere compiled from April 2013 to October 2017 from the following twodata buoys: Station 51,205 (NDMC, 2018a) located 41 km NW of centerof the study region (center: 20.73623°N, 156.69085°W), and Station51,003 (NDBC, 2018b) located 436 km WSW of the center of the studyregion. Final selection of stations 51,205 and 51,003 for analysis wasbased on the availability of continuous data for the entire duration ofthe study period as well as physical location. Before analysis was con-ducted, data were quality controlled by removing missing data, denotedwith variable number of 9's. To assess the impacts of these variables ondaily debris count, weather variables were modelled at the time ofsurvey and the day prior.

Debris retention and accumulation on beaches in Maui is known tobe impacted by ocean factors such as wave and tide height (Blickleyet al., 2016). As such, both land and ocean variables were consideredwhen evaluating potential drivers of land-based debris. For land basedvariables, each survey date was associated with the following datataken from station KLIH1 – 1615680 (National Climatic Data Center's(https://www.ncdc.noaa.gov), 2018) located 26 km NE of the center ofthe study region (center: 20.73623°N, 156.69085°W): average dailywind speed (km/h) and direction (degrees), fastest wind speed (km/h)and direction (deg), and precipitation (Y/N). Ocean-based variableswere taken from Station 51,205 (NDMC, 2018a).

Each month of the survey period was classified by the presence ofthe following: (1) El Nino-Southern Oscillations (ENSO) event, (2) LaNina-Southern Oscillations (LNSO) event, or (3) no event. Data used inanalysis were taken from the NOAA Climate Prediction Center (2018)ENSO monthly categorization table.

Monthly sea surface temperatures from April 2013 to October 2017were downloaded from NOAA Earth System Research LaboratoryPhysical Sciences Division (2018). The 1.0° latitude by 1.0° longitudeglobal grid was loaded into ArcMap (Environmental Systems ResourceInstitute, 2018) and the contour tool in the spatial analyst extensionwas used to create an 18 °C isotherm. The proximity (distance in km) ofthis isotherm to the center of the study region (center: 20.73623°N,156.69085°W) was then calculated in ArcMap. Each month of thesurvey period was then classified by the distance to the 18 °C isotherms.The 18 °C isotherm can be used as an index for the proximity of theSubtropical Convergence Zone (STCZ) (Pichel et al., 2007), with theexpectation that debris loads are higher when STCZ is closer to Hawaii(Ribic et al., 2012).

To determine if tourism influenced land-source debris items, thetotal monthly visitor days from 2013 to 2017 were obtained from theHawaii Tourism Authority (HTA, 2018). Monthly visitor days, calcu-lated by multiplying total monthly visitor count (tourists/month) by theaverage monthly visitor duration (days), ranged from 1.30 million to2.10 million. To facilitate analysis, monthly visitor days were classifiedinto four categories ranging from 1 (lowest number of visitor days) to 4(highest number of visitor days) as follows: 1 (1.30–1.49 million); 2

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

73

(1.50–1.69 million); 3 (1.70–1.89 million); 4 (1.90–2.10 million). All54 months of the survey period were assigned a value of 1 to 4, cor-responding to the appropriate ranges as determined from HTA data set.To facilitate interpretation of results, the average monthly visitor daysand air temperatures (°C) from January 2013 to December 2017 forMaui were summarized and are presented in Appendix Fig. 5.

Given the potential variability in general- and unknown-sourcedebris, a combination of ocean and land-based variables described inprevious sections were used in selecting the best general- and unknown-source models, similar to that of Ribic et al. (2012).

Analysis began with testing of a full candidate model, including allpossible explanatory variables and using AIC to rank the models(Burnham and Anderson, 2002) and determine which variables werecandidates for removal from the model. Variables were then removed ina stepwise manner until a minimum AIC value was reached. Sig-nificance was assessed at α = 0.05 and the minimum AIC models werepresented in the results.

From April 6, 2013 to October 12, 2017, 38,270 km was surveyed(Fig. 2A), and 2095 pieces of marine debris were documented. Marinedebris was observed in all parts of the survey area (Fig. 2B). Debrisspatial quantities (total debris items/km of survey effort/grid cell) overthe total survey period showed a trend of higher accumulation betweenthe islands of Maui, Lanai, and Kahoolawe in the area where the Auau,Kealaikahiki, and Lalakeiki channels meet (Fig. 2B).

Of the 2095 debris items documented, the majority of the debris wasclassified as general-sourced debris (Fig. 3A). Plastics were the pre-dominant type of debris recorded within the study area, accounting for90% of total debris (Fig. 3B).

Quantities of land, ocean, general-source, and unknown-sourcedebris varied between years, with 2017 having the highest quantity ofdebris observed over the five year study period (Table 1). Quantitieswere found to vary between year (Sum Sq: 0.73, F-value: 33.40, p-value: < 0.001) and source category (Sum Sq: 1.58, F-value: 34.42, p-value: < 0.001). Of the debris that could be identified as land or oceanbased, the majority was ocean based; which in some years was fourtimes the concentration of land based debris (Table 1).

The proportion of ocean-based debris was highest in 2013. Therewas a general decreasing trend in general-source debris and a generalincreasing trend of unknown-source debris throughout the surveyperiod. Land-based debris increased with time having highest propor-tion in 2016 and 2017 (Table 1).

An overall increase in weekly debris counts (items/week) was ob-served throughout the study period, with a steep increase fromMarch–May in 2017 (Fig. 4). A 354.5% increase in all debris quantities(items/km effort) was observed in 2017 when compared to average ofthe previous four years, the majority of which was attributed to general-

and unknown-source debris (Table 1). With the exception of 2015, themaximum yearly cluster concentration (number of debris items accu-mulated within 50 m of each other) increased with year (Table 1). Themean cluster size was highest during 2016 and 2017, with maximumdebris cluster in 2017 nearly double the average of the previous fouryears (Table 1).

The majority of plastic debris items consisted of plastic (36%) andfoam (14%) fragments (Fig. 5A). For items that were found whole,plastic bottles (16%) and buoys (14%) accounted for nearly one-third ofthese collected (Fig. 5B).

Of the debris documented, 73.3% (n = 1536) exhibited some formof biofouling, with plastics comprising the largest proportion(n = 1425, 92.7%) of biofouled items. The amount of biofouling variedby item, but was highest for buckets (avg = 55.8%) and lowest forballoons (avg = 1.4%) (Fig. 6). Eight items contained biofouling notnative to Hawaiian waters, including blue mussels (Mytilus edulis),chitons (Mopalia), and/or limpets (Lottia), which were initially identi-fied in the field before being photographed and sent to the Departmentof Aquatic Resources for confirmation when possible. Foreign writingallowed for assessment of probable country of origin for 23 items. Ofthese items, 10 items displayed Japanese characters, 7 displayed Chi-nese characters, and 4 displayed Korean characters. Two items dis-played characters which could have belonged to either the Japanese orChinese languages, and therefore could not be identified to a country.The majority of these items (13 items, 56.5%) were identified as comingfrom ocean sources, while the remaining 10 items could have originatedfrom either land- or ocean-based sources.

The subset of 252 hard plastic debris items collected from April 14to July 11, 2017 was analyzed to determine polymer composition. Themajority (52.5%) of debris items analyzed was classified as severelyweathered, with the remaining items being mildly (27%) and moder-ately (25.5%) weathered. Weathering code did not impact polymercomposition, because pieces were cut to reveal an inner surface for ATRFT-IR measurements. All pieces consisted of polymers that would floatin seawater, based on the polymer density, except for one PET bottle.HDPE, LDPE, and polypropylene (PP) accounted for the largest pro-portion of debris sampled (Fig. 4a & b). PP makes up the greatestproportion of sampled debris by mass, while HDPE and LDPE make upthe greatest proportion by count (Fig. 7A & B).

The lowest AIC model for the entire dataset, regardless of debrissource category, included ocean-based sets of variables/processes. Datafrom Buoy 51,003 with a one day offset resulted in the best fit model forthis dataset. The most significant driver of debris counts (items/day)was coded survey day (Table 2; Fig. 8B). Additional factors, in de-creasing order of significance included: the interaction between waveheight (m) and dominant direction (degrees) (Table 2; Fig. 9), non-

Fig. 3. Proportions of debris (A) origin and (B) material collected between April 6, 2013 and October 12, 2017 within the coastal waters of Maui County, Hawaii.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

74

ENSO/LNSO months (Table 2), and distance to the 18 °C isotherm(Table 2; Fig. 8A). The combination of low wave height (< 1 m) andwind coming from 150° or 350° resulted in high debris counts (Fig. 9).Debris counts fluctuated with distance to the 18 °C isotherm, but therewas a general decreasing trend of debris with increasing distance to the18 °C isotherm (Fig. 8A). Strong temporal trends in debris counts wereevident with peaks observed in April 2015 and February 2017, each ofwhich were preceded by dips in debris counts. A significant increase indebris counts was observed during non-ENSO/LNSO months (Table 2).

The lowest AIC model for ocean-based debris was fit using variablesfrom Buoy 51,003, with a 1 day offset. Ocean-based debris counts(items/day) showed significant nonlinear relationships with wind di-rection, air temperature, and survey date; and significant linear re-lationships with sea level pressure and wave period (Table 3; Fig. 10).Air temperature could indicate an influence of the STCZ, as tempera-tures would be coldest in late winter and early spring when the zone is

closest to Hawaii. As such, air temperature may represent a lag effect ofthe STCZ with high accumulation being observed after the zone reachesits closest point to Maui. However, further research is needed to de-termine the exact lag-time and potential connection between STCZ andair temperature.

Ocean-based debris counts remained fairly constant until June2016, which saw the sharpest increase in debris until December 2016followed by the sharpest decrease in debris until October 2017(Fig. 10B). Peaks in ocean-based debris counts varied based on winddirections, with peaks occurring when wind direction was 0, 75, 125,and 200° (Fig. 10C).

Land-based debris counts showed significant nonlinear relationshipswith water temperature and average wind speed (Table 4; Fig. 11).Land-based debris counts were highest during low wind speed intervals(6–10 km/h) and high speed intervals (22–24 km/h), with variationsfrom 6 to 12 mph (Fig. 11A). There were minimal changes to land-based

Table 1Yearly quantities, proportions, and cluster sizes of marine debris items summarized by total and source categories documented between April 6, 2013 and October 12,2017 within the coastal waters of Maui County, Hawaii.

Year 2013 2014 2015 2016 2017

Debris summaryTotal debris count 402 276 236 328 853Total survey effort (km) 5985 7963 5067 4218 4248Total survey days 47 62 43 36 31

Quantities of debrisQuantities of all debris (items/km effort) 0.067 0.036 0.048 0.080 0.201Quantities of ocean-based debris (items/km effort) 0.013 0.006 0.008 0.011 0.009Quantities of land-based debris (items/km effort) 0.003 0.003 0.004 0.009 0.009Quantities of general-source debris (items/km effort) 0.036 0.018 0.027 0.033 0.097Quantities of unknown debris (items/km effort) 0.015 0.008 0.008 0.027 0.085

Proportions of debrisProportion of ocean-based debris 19.40 17.54 17.43 14.29 4.69Proportion of land-based debris 4.44 8.07 9.13 10.71 4.34Proportion of general-source debris 53.73 50.88 56.02 41.37 48.53Proportion of unknown debris 22.39 23.51 17.43 33.63 42.44

Debris clustersMean items/clustera (standard deviation) 1.52 (1.49) 1.43 (1.49) 1.20 (0.88) 1.80 (2.10) 2.21 (3.06)Maximum items/clustera 10 13 8 14 21

a Debris items located within 50 m of each other were considered part of the same cluster.

Fig. 4. Weekly counts (items/week) of debris items collected between April 6, 2013 and October 12, 2017 within the coastal waters of Maui County, Hawaii.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

75

debris accumulation on survey days when air temperatures ranged from24.0 to 25.0 °C, with a gradual reduction in debris seen when tem-peratures exceeded 25.0 °C (Fig. 11B). The months with the highesttemperatures correspond to months with lowest monthly visitor days:August, September, and October (Appendix Table 2).

The lowest AIC model for general-source debris was fit using vari-ables from Buoy 51,003, with a 1 day offset; the same as the ocean-based model. General-source debris counts showed significant non-linear relationships with survey date, and marginally significant linearrelationships with year (Table 5; Fig. 12). General-source debris countsgradually declined throughout the survey period (Fig. 12). Althoughnot significant, within year counts showed an increasing positive linearrelationship with increasing years (Table 5).

The lowest AIC model for unknown-source debris was fit usingvariables from Buoy 51,003, with a 1 day offset; the same as the ocean-based and general-source debris models. Unknown-source debris countsshowed significant nonlinear relationships with water temperature, and

significant linear relationships with survey date, ENSO, wind speed,and peak gusts (Table 6; Fig. 13). Unknown-source debris countsshowed a varying trend with temperature (Fig. 13). Significant positivetrends of unknown-source debris counts with survey date, southernoscillations, and wind speed were observed, while a negative lineartrend was found with increasing peak gusts (Table 6).

The observed variation in significant variables based on marinedebris source category aligns with work presented in Ribic et al. (2012).Overall, the difference in the sets of variables included in the lowest AICmodel for each source category and for all debris, regardless of source,suggests a wide variety of drivers are likely responsible for the varia-bility in debris quantities observed in the nearshore waters of MauiCounty. The prevalence of survey date and ocean based variables with aone day offset in four of the five lowest AIC models suggests theseprocesses are linked to temporal and ocean based drivers and are bestcaptured using time and an offshore buoy data set (Buoy 51,003).

As reported in Currie et al. (2017), plastics comprised the majority

Fig. 5. Proportions of (A) plastic debris items (n = 1986) divided into subcategories and (B) intact debris items (n = 887) divided into commonly sighted sub-categories groups collected between April 6, 2013 and October 12, 2017 within the coastal waters of Maui County, Hawaii.Note: Panels A and B were created separately and the same item may be used in both figures. As such, these figures should be evaluated independently.

0102030405060708090100

)%(

egarevoCluofoi

B

Groups

Max/Min Average

Fig. 6. Average percent biofouling observed on whole debris items divided into identifiable groups collected between April 6, 2013 and October 12, 2017 within thecoastal waters of Maui County, Hawaii.Note: Error bars represent the standard deviations.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

76

of floating macro debris found in this study region; a result that alignswith the known prevalence of floating plastics in the ocean (Coe andRogers, 1997; Derraik, 2002). The increasing trend in debris quantitiesthroughout the duration of the study aligns with the global trend ofincreasing debris deposition in our oceans (Erikssen et al., 2014). Thelarge increase in debris quantities observed in 2017 is likely related tothe higher number of small scale debris clusters, which were observedmore frequently and in larger sizes in 2017. Although observerschanged throughout the survey period, the number of observers re-mained constant. As such, it is unlikely that differences in observersaccounted for the substantial increase in debris observed in 2017. Si-milarly, weather conditions were also kept consistent (BSS andDSS ≤ 3) throughout the study period and weather changes likely donot account for the observed increase.

The predominantly northwest surface currents in leeward areas ofMaui County occur, in part, from Ekman transport along with wind-driven eddy effects resulting from the northeast trade winds interacting

with the land masses of the islands (Chavanne et al., 2002). These ed-dies likely result in the convergence pattern of debris seen in thechannels that separate the four islands of the region; opposing eddies inthe lee of Maui Island may cause areas of lower current velocities,which allows marine debris to accumulate.

Debris from Japan, China, and Korea most likely drifted to theHawaiian Islands after transport in the Subtropical Gyre and subsequentnorthwest surface currents toward the leeward waters of Maui (e.g.,Chavanne et al., 2002; Howell et al., 2012). Although plausible countryof origin was determined via markings on the debris, it is virtuallyimpossible to determine the exact point at which any particular itementered the marine environment. For example, debris with Japanesewriting may have entered the ocean in coastal Japan, from an offshorefishing vessel, or from a tourist visiting Hawaii.

The composition of debris presented here supports previous reportsthat the majority of marine debris in and around the Hawaiian Islandsoriginates from far offshore rather than local land-based sources

Fig. 7. Percent of polymers identified in debris sub-sample (n = 252) as a function of total (A) count and (B) mass. Low-density polymers expected to float onseawater are shown in white and blues.Note: Polyethylene terephthalate (PET) (n = 1), high-density polyethylene (HDPE), low-density polyethylene (LDPE), other PE is a piece that had an obvious PEspectrum but with additional non-PE peaks (n = 1), polyethylene and polypropylene mixture (PE/PP mix) as defined in Jung et al. (2018), polypropylene (PP),ethylene vinyl acetate (EVA), and unidentifiable (n = 1) were pieces that produced very noisy spectrum that could not be identified.

Table 2Results of top generalized additive model used for determining the linear and nonlinear relationships between all debris counts and variables, based on data collectedwithin the coastal waters of Maui County, Hawaii between 2013 and 2017.

Factor edfa F-value p-Value R2 Dev. expl.

All debris nonlinear s(18 °C isotherm) 8.22 1.95 0.05 0.45 52.1%s(wave height, dominant wave direction) 9.69 2.61 0.003s(survey date) 7.60 7.57 < 0.0001

Factor Estimate t-Value p-Value R2 Dev. expl.

All debris linear La Nina Oscillation 0.52 0.81 0.42 0.45 52.1%No Oscillation 1.35 2.21 0.02

a edf is the estimated degrees of freedom accounting for the smoothing function.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

77

(Donohue et al., 2001; Ribic et al., 2012). The proportion of general-source debris recorded in this study is slightly higher than the 30–40%recorded in other shoreline surveys in Hawaii (Ribic et al., 2012). Thisis likely attributed to the addition of items to the general-source cate-gory for this study, not included in the original designation by Ribicet al. (2012). Furthermore, the high proportion of unknown-sourcedebris and the observed biofouling suggests that few items were litteredinto the environment recently, and thus could have come from verydistant locations.

The subset of samples analyzed in 2017 suggests that only low-density floating debris were observed in the Maui County region. Thehigh proportion of severely weathered debris suggests that few itemswere littered into the environment recently and most come from distantsources. These results are congruent with previous studies that found PEand PP to dominate sea surface plastic marine debris in the NorthPacific (Brandon et al., 2016), North Atlantic (ter Halle et al., 2016),Indian Ocean (Syakti et al., 2017), Mediterranean Sea (Pedrotti et al.,2016), Ross Sea (Cincinelli et al., 2017), coastal Southern Malaysia (Ngand Obbard, 2006), and on beaches of Kauai, Hawaii (Cooper andCorcoran 2010). Any debris made of PET, PVC, PS, nylon, and otherdenser polymers entering Hawaiian waters from local sources wouldsink and the collection of only visible floating debris for this studyexplains the near absence of this type of polymer in the analysis. Assuch, a large amount of the marine debris is likely going undetected, asit is sinking through the water column or on the sea floor. The one pieceof PET plastic collected and analyzed during the survey would havenaturally sunk, but the item was a bottle that still had air inside, whichkept it afloat until sample collection. Had it filled with water, it wouldhave certainly sank and been deposited on the sea floor.

As has been shown in previous work by Ribic et al. (2010, 2011,

2012), the complex relationships of debris accumulation and varyingdrivers leads to temporal patterns of debris accumulation. Blickley et al.(2016) found debris accumulation on Maui's beaches fluctuated on amonthly and daily basis and resuspension of debris was related to wind,tides and wave height. Ocean based phenomena such as ENSO events,as well as proximity to the STCZ, as indicated by the 18 °C isotherm(Pichel et al., 2007), did have a significant impact on overall debrisquantities and unknown-source debris observed in this study. Althoughnot assessed in this study, there could be a delayed pulse of debris ac-cumulation occurring after strong El Nino years, explaining the sig-nificance positive debris counts during non ENSO event months ob-served in this study. For example, the 2016 moderate to strong El Ninocould have caused the high accumulation event observed in 2017, andnot be predicted by ENSO months or proximity to the STCZ used in ourmodels due to an untested lag effect. This movement of the STCZ fur-ther south and closer to Hawaii could have resulted in deposition ofhigh amounts of debris into Hawaii's dynamic coastal waters. This mayhave caused the increase in debris accumulation observed in 2017, asthe debris was subject to movement through this dynamic system. Therole of ENSO events and the STCZ on coastal debris accumulationwarrants additional research to help understand the potential connec-tion to high accumulation events.

Significant variables for all debris sources were a mix of local andlarge scale phenomena, suggesting a complex process of drivers, mostlyrelating to ocean based variables, are responsible for the observedvariations in debris quantities within Maui County. There were fivevariables identified as significant drivers of ocean-based debris, sug-gesting multiple factors lead to the accumulation of ocean-based debriswithin Maui County. The low R2 value suggests other untested factorsare contributing to ocean-based debris fluctuations. The increase inocean-based debris count with temperature could be contributed to theproximity of the STCZ to Maui, which is closer during colder periodsand is known to concentrate debris (Pichel et al., 2007). This is furthersupported by the significance of the 18 °C isotherm in increasing anddecreasing all debris counts regardless of source, suggesting that ob-served debris likely originate outside the Hawaiian Islands. Tempera-ture could also serve as a proxy for various oceanic processes (Ryanet al., 2009) that were not specifically tested here. The large fluctua-tions in ocean-based debris over time observed between 2015 and 2016aligned with one of the strongest El Niño events observed since 1950(ENSO, 2016). The absence of the ENSO variable from the lowest AICmodel for the ocean-based debris likely resulted from debris that ar-rived via the ocean being classified as unknown-source due to inabilityto identify the source. Additionally, interaction effects were not testedand only considering the ENSO variable on its own could further ex-plain the absence from this model. This is further supported by theinclusion of the ENSO variable in the model looking at all debris re-gardless of source category.

The general increase in ocean-based debris from 100 to 250° cor-responds to wind coming from a direction unobstructed by any of thefour islands, with the peak of 200–250° (from Southwest) representingthe least obstructed area between Kahoolawe and Lanai.

The increase in sea level pressure read at the buoy southwest of thestudy area is indicative of a shift in the high pressure ridge over theislands, which causes a stalling of the prevailing Northeast trade windscommon to the Hawaiian Islands (Garza et al., 2012). The observedincrease in ocean-based debris quantities with increasing sea levelpressure suggests this stalling of trade winds slows the transport ofocean-based debris out of the study area. This is further supported bythe significant increase in debris observed during non-ENSO/LNSOmonths for all debris, as this is when a slowing of the trade winds isexpected.

The general decrease in land-based debris with water temperatureobserved could be attributed to seasonal beach use by residents andtourists, not detected using visitor count categories. Sea surface tem-peratures are hottest in September and October (National Data Buoy

Fig. 8. Results of generalized additive model showing the significant non-linearrelationships of all debris and (A) distance to 18 °C isotherm and (B) survey datefor debris collected within coastal waters of Maui County, Hawaii between2013 and 2017.Note: Survey date is coded so that 1 = April 6, 2013 and 1634 = October 12,2017.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

78

Center 2018a), which corresponds with the lowest monthly averagevisitor days for the island of Maui (Hawaii Tourism Authority, 2018)and could explain the reduction in debris counts from 25 to 27 °C.Temperatures in the range of 24–25 °C correspond to the peak tourismmonths of December to March which may explain the peak observed inland-based debris at these temperatures. The authors believe watertemperature in Hawaii could be an inverse measure to indicate thenumber of beach users. However, further research is needed to confirmthese trends and to assess if tourists contribute more or less marinedebris than local residents. Population levels have been shown to im-pact debris deposition (Thiel et al., 2011) and variations in visitors toMaui may impact the amount of debris observed through direct de-position and/or deposition via transport systems such as streams to theocean system (Ribic et al., 2012).

Wind is a primary driver of debris transport over land and deposi-tion into coastal waters (Blickley et al., 2016). The initial trend of highland-based debris deposition with low wind speeds could relate to thenumber of beach users because calm, windless days are favored for

beach activities, which are common throughout Maui. The variabilityobserved in land-based debris deposition with wind speed is not sur-prising as the amount and rate of land-based debris collection in thisstudy was likely the result of complex drivers that vary with beach site.Blickley et al. (2016) showed that land-based accumulation on beacheswas mostly influenced by wind, which was also the most significantterm for accumulation of land-based debris in Maui's nearshore waters.

General-source debris represented the largest proportion of datacollected in this study and can be affected by land-based or ocean-basedprocesses (Barnes et al., 2009; Ribic et al. 2012). Survey date was theonly variable that was significant in the general-source model andcommon with the lowest AIC model for ocean-based debris. This sug-gests that general-source debris accumulation in Maui County has atemporal component, which may be linked to ocean processes as sug-gested in Ribic et al. (2012).

The decreasing non-linear trend in general-source debris countsfrom 2013 to 2017 is worth noting as it corresponds with increasinglinear estimates of yearly debris counts. Although the yearly estimates

Fig. 9. Results of generalized additive model showing the significant relationship of the interaction between wave height and direction on all debris counts collectedwithin coastal waters of Maui County, Hawaii between 2013 and 2017.

Table 3Results of top generalized additive model used for determining the linear and nonlinear relationships between ocean-based debris counts and variables, based on datacollected within the coastal waters of Maui County, Hawaii between 2013 and 2017.

Factor edfa F-value p-Value R2 Dev. expl.

Ocean-based nonlinear s(wind direction) 7.86 2.41 0.02 0.14 43.6%s(air temp) 3.28 5.70 < 0.0001s(survey date) 7.27 3.21 < 0.001

Factor Estimate t-Value p-Value R2 Dev. expl.

Ocean-based linear Wave height 0.09 1.37 0.17 0.14 43.6%Dominant wave period −0.06 −2.13 0.03Sea level pressure 0.13 3.26 < 0.001

a edf is the estimated degrees of freedom accounting for the smoothing function.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

79

are not significant, these results would suggest a strong temporalcomponent that varies on a daily and yearly basis and is likely a proxyfor a process not tested in our models.

There were six variables identified as significant drivers of un-known-source debris, all relating to ocean-based environmental vari-ables suggesting the majority of debris in this category has been floatingin the ocean for prolonged periods of time. This is further supported bythe fragmented nature of debris in this category, likely being brokendown from sun, wind, and wave action. The variation in unknown-source debris counts with water temperature is likely the result of themovement of the STCZ to Maui, which is known to concentrate debris(Pichel et al., 2007). The high significance counts of unknown-debrisduring non ENSO event months is similar to what was observed for alldebris and likely is the result of a similar process and a delayed pulseafter a strong ENSO month. The increasing trend of unknown-debriscounts with increasing wind speed and the decreasing trend of un-known-debris counts with increasing peak gusts is likely attributed to

the topography of the Maui Nui region. Further, it suggests unknown-debris is largely influenced by wind driven currents and waves withincreasing wind speed increasing accumulation and high gusts movingdebris either onshore, or beyond the study region.

The use of an unknown-source category represents an importantaddition to the categories presented by Ribic et al. (2012), for theHawaii region as it represents a large portion of the observed debris.The expansion of categories proposed by Ribic et al. (2012) to includean unknown-source of debris fragments is further strengthened by thefact that plastic fragments make up an estimated 96% of the plasticsfound in the North Pacific (Robards et al., 1997). As such, identifyingsignificant drivers for this category will contribute to a better under-standing of debris accumulation in the North Pacific.

To mitigate the problem of marine debris an understanding of howit gets into our environment is required. To the best of the author'sknowledge, this is the first published study in Hawaii to conduct sys-tematic ocean surveys as a method of quantifying marine debris and, as

Fig. 10. Results of generalized additive model showing the significant non-linear relationships of ocean-based debris and (A) air temperature, (B) survey date, and(C) wind direction for debris documented within coastal waters of Maui County, Hawaii between 2013 and 2017.Note: Survey date is coded so that 1 = April 6, 2013 and 1634 = October 12, 2017. Fig. 10C represents trends over recorded over the dominant wind direction, whichranged from 35 to 264° throughout the survey period.

Table 4Results of top generalized additive model used for determining the linear and nonlinear relationships between land-based debris counts and variables, based on datacollected within the coastal waters of Maui County, Hawaii between 2013 and 2017.

Factor edfa F-value p-Value R2 Dev. expl.

Land-based nonlinear s(wave height) 1.82 0.85 0.47 0.32 57.1%s(average wave period) 2.80 2.310 0.07s(water temperature) 2.19 3.71 0.02s(average wind speed) 7.02 2.89 0.01

Factor Estimate t-Value p-Value R2 Dev. expl.

Land-based linear Dominant wave direction −0.01 −1.33 0.19 0.32 57.1%

a edf is the estimated degrees of freedom accounting for the smoothing function.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

80

such, provides valuable baseline information on the sources and accu-mulation patterns of pollution at the sea surface in this region. Removalefforts are useful in getting debris out of the environment, but quanti-fying the types, sources, and amounts of debris is essential to stoppingthis problem at the source. Systematic research is needed at the regionallevel because each area will have its own unique drivers and trends.Citizen science programs, such as Pacific Whale Foundation's CoastalMarine Debris Monitoring Program, can serve as a low cost method ofobtaining data so researchers can monitor debris types and loads.

Baseline data are important to obtain so that trends can be detected andthese data are crucial for managers and lawmakers to implement in-formed, scientifically-backed policies and mitigation measures.

Disclaimer: Certain commercial equipment, instruments, or mate-rials are identified in this paper to specify adequately the experimentalprocedure. Such identification does not imply recommendation or en-dorsement by the National Institute of Standards and Technology, nordoes it imply that the materials or equipment identified are necessarilythe best available for the purpose.

Fig. 11. Results of generalized additive model showing the significant non-linear relationships of land-based debris and (A) wind speed and (B) sea surface tem-perature for debris documented within coastal waters of Maui County, Hawaii between 2013 and 2017.

Table 5Results of top generalized additive model used for determining the linear and nonlinear relationships between general-source debris counts and variables, based ondata collected within the coastal waters of Maui County, Hawaii between 2013 and 2017.

Factor edfa F-value p-Value R2 Dev. expl.

General nonlinear s(month) 6.83 1.66 0.12 0.30 43.1%s(survey date) 5.29 6.26 < 0.001s(water temperature) 6.95 8.05 0.11

Factor Estimate t-Value p-Value R2 Dev. expl.

General linear Year 2014 3.95 1.49 0.14 0.30 43.1%Year 2015 9.52 1.80 0.07Year 2016 14.41 1.83 0.07Year 2017 19.48 1.85 0.06

a edf is the estimated degrees of freedom accounting for the smoothing function.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

81

Fig. 12. Results of generalized additive model showing the significant non-linear relationships of general-source debris and survey date for debris collected withincoastal waters of Maui County, Hawaii between 2013 and 2017.Note: Survey date is coded so that 1 = April 6, 2013 and 1634 = October 12, 2017.

Table 6Results of top generalized additive model used for determining the linear and nonlinear relationships between unknown-source debris counts and variables, based ondata collected within the coastal waters of Maui County, Hawaii between 2013 and 2017.

Factor edfa F-value p-Value R2 Dev. expl.

Fragments nonlinear s(average wave period) 1.27 1.51 0.17 0.45 51.5%s(water temperature) 4.41 3.92 < 0.001

Factor Estimate t-Value p-Value R2 Dev. expl.

Fragments linear Survey date 0.001 7.37 < 0.0001 0.45 51.5%La Nina Oscillation 0.682 2.33 < 0.01No Oscillation 1.01 4.81 < 0.0001Average wind speed 1.46 2.36 < 0.01Peak gusts −1.32 −2.44 < 0.01

a edf is the estimated degrees of freedom accounting for the smoothing function.

Fig. 13. Results of generalized additive model showing the significant non-linear relationships of unknown-source debris and water temperature for debris docu-mented within coastal waters of Maui County, Hawaii between 2013 and 2017.Note: Survey date is coded so that 1 = April 6, 2013 and 1634 = October 12, 2017.

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

82

Acknowledgements

The authors would like to thank Jessica McCordic, Grace Olson, andAbigail Machernis for help with data collection and quality control andthe Pacific Whale Foundation research interns who helped with datacollection and input from 2013 to 2017. We would also like to thankJared Ragland and Mark Manual for their guidance and comments onthe initial review of this manuscript. Special thanks goes to RachaelLallo for creating several appendix figures. We thank Pacific WhaleFoundation's members and supporters for providing the funding ne-cessary to conduct this study.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.marpolbul.2018.11.026.

References

Barnes, D.K.A., Galgani, F., Thompson, R.C., Barlaz, M., 2009. Accumulation and frag-mentation of plastic debris in global environments. Philos. Trans. R. Soc. 364,1985–1998. https://doi.org/10.1098/rstb.2008.0205.

Blickley, L.C., Currie, J.J., Kaufman, G.D., 2016. Trends and drivers of debris accumu-lation on Maui shorelines: implications for local mitigation strategies. Mar. Pollut.Bull. 105, 292–298.

Brandon, J., Goldstein, M., Ohman, M.D., 2016. Long-term aging and degradation ofmicroplastic particles: comparing in situ oceanic and experimental weathering pat-terns. Mar. Pollut. Bull. 110 (1), 299–308.

Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference, aPractical Information – Theoretic Approach, 2nd ed. Springer-Verlag, New York(512 p).

Chavanne, C., Flament, P., Lumpkin, R., Dousset, B., Bentamy, A., 2002. Scatterometerobservations of wind variations induced by oceanic islands: implications for wind-driven ocean circulation. Can. J. Remote. Sens. 28 (3), 466–474.

Cincinelli, A., Scopetani, C., Chelazzi, D., Lombardini, E., Martellini, T., Katsoyiannis, A.,Fossi, M.C., Corsolini, S., 2017. Microplastic in the surface waters of the Ross Sea(Antarctica): occurrence, distribution and characterization by FTIR. Chemosphere175, 391–400.

Coe, J.M., Rogers, D.B., 1997. Marine Debris: Sources, Impacts and Solutions. Springer-Verlag, New York, NY, pp. 49–66.

Cooper, D.A., Corcoran, P.L., 2010. Effects of mechanical and chemical processes on thedegradation of plastic beach debris on the island of Kauai, Hawaii. Mar. Pollut. Bull.60 (5), 650–654.

Currie, J.J., Stack, S.H., McCordic, J.A., Kaufman, G.D., 2017. Quantifying the risk thatmarine debris poses to cetaceans in coastal waters of the 4-island region of Maui.Mar. Pollut. Bull. 121 (1–2), 69–77.

Dameron, O.J., Parke, M., Albins, M.A., Brainard, R., 2007. Marine debris accumulationin the Northwestern Hawaiian Islands: an examination of rates and processes. Mar.Pollut. Bull. 54 (4), 423–433.

Derraik, J.G.B., 2002. The pollution of the marine environment by plastic debris: a re-view. Mar. Pollut. Bull. 44, 842–852.

Donohue, M.J., Boland, R.C., Sramek, C.M., Antonelis, G.A., 2001. Derelict fishing gear inthe Northwestern Hawaiian Islands: diving surveys and debris removal in 1999confirm threat to coral reef ecosystems. Mar. Pollut. Bull. 42 (12), 1301–1312.

Erikssen, M., Lebreton, L.C.M., Carson, H.S., Thiel, M., Moore, C.J., Borerro, J.C., Galgani,F., Ryan, P.G., Reisser, J., 2014. Plastic pollution in the world’s oceans: more than 5trillion plastic pieces weighing over 250,000 tons afloat at sea. PLoS One 9, 1–15.

ENSO, 2016. Pacific ENSO update. A quarterly bulletin of the Pacific El Nino-SouthernOscillations Applications Climate (PEAC) Center. 22 (2), 1–16.

Environmental Systems Research Institute, 2012. ArcGIS Desktop: Release 10.4.Environmental Systems Research Institute, Redlands, CA.

Escardó-Boomsma, J., O'Hara, K., Ribic, C.A., 1995. A National Marine Debris MonitoringProgram, Final Report. US Environmental Protection Agency, Office of Water,Washington, DC (40p).

Fox, J., Weisberg, S., 2011. An R Companion to Applied Regression, Second edition. SAGEPublications, Inc., California (472 p).

Galloway, T.S., Cole, M., Lewis, C., 2017. Interactions of microplastic debris throughoutthe marine ecosystem. Nat. Ecol. Evol. 1 (5), 0116.

Garza, J.A., Chu, P.S., Norton, C.W., Schroeder, T.A., 2012. Changes of the prevailingtrade winds over the islands of Hawaii and the North Pacific. J. Geophys. Res. Atmos.117 (D11), 1–18.

Goldstein, M.C., Titmus, A.J., Ford, M., 2013. Scales of spatial heterogeneity of plasticmarine debris in the Northeast Pacific Ocean. PLoS One 8 (11), 1–11.

Gonzalez-Suarez, M., Gómez, A., Revilla, E., 2013. Which intrinsic traits predict vulner-ability to extinction depends on the actual threatening processes. Ecosphere 4 (6),1–16.

Grigg, R., Grossman, E., Earle, S., Gittings, S., Lott, D., McDonough, J., 2002. Drownedreefs and antecedent karst topography, Auau Channel, SE Hawaiian Islands. Coral

Reefs 21, 73–82.Hawaii Tourism Authority, 2018. Historical visitor statistics. http://www.

Hawaiitourismauthority.org/research/reports/, Accessed date: April 2018.Howell, E.A., Bograd, S.J., Morishige, C., Seki, M.P., Polovina, J.J., 2012. On North

Pacific circulation and associated marine debris concentration. Mar. Pollut. Bull. 65(1–3), 16–22.

Jung, M.R., Horgen, F.D., Orski, S.V., Rodriguez, C.V., Beers, K.L., Balazs, G.H., Jones,T.T., Work, T.M., Brignac, K.C., Royer, S.-J., Hyrenbach, K.D., Jensen, B.A., Lynch,J.M., 2018. Validation of ATR FT-IR to identify polymers of plastic marine debris,including those ingested by marine organisms. Mar. Pollut. Bull. 127, 704–716.https://doi.org/10.1016/j.marpolbul.2017.12.061.

Kubota, M., 1994. A mechanism for the accumulation of floating marine debris north ofHawaii. J. Phys. Oceanogr. 24, 1059–1064.

Kühn, S., Rebolledo, E.L.B., van Franeker, J.A., 2015. Deleterious effects of litter onmarine life. In: Marine Anthropogenic Litter. Springer, Cham, pp. 75–116.

Mann, J., 1999. Behavioral sampling methods for cetaceans: a review and critique. Mar.Mamm. Sci. 15 (1), 102–122.

Moore, C.J., 2008. Synthetic polymers in the marine environment: a rapidly increasing,long-term threat. Environ. Res. 108 (2), 131–139.

National Climate Data Center (NCDC), 2018. NOAA Automated Weather ObservingSystem (AWOS) Stations. National Climatic Data Center, Ashville, North Carolina,USA (Data Accessed April 2018).

National Data Buoy Center, 2018a. Standard Meteorological Data Collected From theNational Data Buoy Center From 2013–2017 (Station 51205 - Pauwela, Maui, HI[187]). Pacific Islands Ocean Observing System (PacIOOS) (Data provided by ScrippsInstitution of Oceanography. Accessed: April 2018).

National Data Buoy Center, 2018b. Standard Meteorological Data Collected From theNational Data Buoy Center From 2013–2017 (Station 51003 (LLNR 28007) -WESTERN HAWAII - 205 NM SW of Honolulu, HI). (Owned and maintained byNational Data Buoy Center. Accessed: April 2018).

Ng, K.L., Obbard, J.P., 2006. Prevalence of microplastics in Singapore's coastal marineenvironment. Mar. Pollut. Bull. 52 (7), 761–767.

NOAA Climate Prediction Center, 2018. Cold and warm episodes by season. http://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php, Accesseddate: April 2018.

NOAA Earth System Research Laboratory Physical Sciences Division, 2018. PSD Griddedclimate datasets: all. http://www.esrl.noaa.gov/psd/data/gridded, Accessed date:April 2018.

Pedrotti, M.L., Petit, S., Elineau, A., Bruzaud, S., Crebassa, J.C., Dumontet, B., Martí, E.,Gorsky, G., Cózar, A., 2016. Changes in the floating plastic pollution of theMediterranean Sea in relation to the distance to land. PLoS One 11 (8), e0161581.

Pichel, W.G., Churnside, J.H., Veenstra, T.S., Foley, D.G., Friedman, K.S., Brainard, R.E.,Nicoll, J.B., Zheng, Q., Clemente-Colón, P., 2007. Marine debris collects within theNorth Subtropical Convergence Zone. Mar. Pollut. Bull. 54, 1207–1211. https://doi.org/10.1016/marpolbul.2007.04.010.

R Core Team, 2017. R: a language and environment for statistical computing. RFoundation for Statistical Computing, Vienna, Austria URL. https://www.R-project.org/.

Ribic, C.A., Sheavly, S.B., Rugg, D.J., Erdmann, E.S., 2010. Trends and drivers of marinedebris on the Atlantic Coast of the United States 1997–2007. Mar. Pollut. Bull. 60,1231–1242. https://doi.org/10.1016/j.marpolbul.2010.03.021.

Ribic, C.A., Sheavly, S.B., Rugg, D.J., 2011. Trends in marine debris in the U.S. Caribbeanand the Gulf of Mexico 1996–2003. J. Integr. Coast. Zone Manag. 11, 7–19.

Ribic, C.A., Sheavly, S.B., Rugg, D.J., Erdmann, E.S., 2012. Trends in marine debris alongthe US Pacific Coast and Hawaii 1998–2007. Mar. Pollut. Bull. 64 (5), 994–1004.

Robards, M.D., Gould, P., Platt, J., 1997. The highest global concentrations and increasedabundance of oceanic plastic debris in the North Pacific: Evidence from seabirds (pp71–80). In: Marine Debris: Sources, Impact and Solutions. Springer, New York.

Rochman, C.M., Hoh, E., Kurobe, T., Teh, S.J., 2013. Ingested plastic transfers hazardouschemicals to fish and induces hepatic stress. Sci. Rep. 3, 3263.

Ryan, P.G., Moore, C.J., van Franeker, J.A., Moloney, C.L., 2009. Monitoring the abun-dance of plastic debris in the marine environment. Philos. Trans. R. Soc. B 364,1999–2012.

Sheavly, S.B., Register, K.M., 2007. Marine debris and plastics: environmental concerns,sources, impacts and solutions. J. Polym. Environ. 15 (4), 301–305.

Syakti, A.D., Bouhroum, R., Hidayati, N.V., Koenawan, C.J., Boulkamh, A., Sulistyo, I.,Lebarillier, S., Akhlus, S., Doumenq, P., Wong-Wah-Chung, P., 2017. Beach macro-litter monitoring and floating microplastic in a coastal area of Indonesia. Mar. Pollut.Bull. 122 (1–2), 217–225.

Ter Halle, A., Ladirat, L., Gendre, X., Goudouneche, D., Pusineri, C., Routaboul, C.,Tenailleau, C., Duployer, B., Perez, E., 2016. Understanding the fragmentation pat-tern of marine plastic debris. Environ. Sci. Technol. 50 (11), 5668–5675.

Thiel, M., Bravo, M., Hinojosa, I.A., Luna, G., Miranda, L., Nunez, P., Pacheco, A.S.,Vasquez, N., 2011. Anthropogenic litter in the SE Pacific: an overview of the problemand possible solutions. J. Integr. Coast. Zone Manag. 11, 115–134.

Van Sebille, E., 2015. The oceans' accumulating plastic garbage. Phys. Today 68, 60–61.Wakata, Y., Sugimori, Y., 1990. Lagrangian motions and global density distributions of

floating matter in the ocean simulated using shipdrift data. J. Phys. Oceanogr. 20 (1),15–28.

Wood, S.N., 2006. Generalized Additive Models. Chapman and Hall, New York (496 p).Wood, S.N., 2017. Generalized Additive Models: An Introduction With R, 2nd ed.

Chapman and Hall/CRC, New York (485 p).Zar, J.H., 1984. Biostatistical Analysis. Prentice-Hall, New Jersey (960 p).

J.J. Currie et al. Marine Pollution Bulletin 138 (2019) 70–83

83